Automatic selection of IMFs to denoise the sEMG signals using EMD

Area of Science:

• Biomedical Engineering
• Signal Processing
• Rehabilitation Technology

Background:

• Surface Electromyography (sEMG) signals are crucial for muscle activity analysis but are susceptible to noise like power line interference and motion artifacts.
• Existing denoising methods like Empirical Mode Decomposition (EMD) and its variants (EEMD, CEEMD) rely on statistical approaches for selecting intrinsic mode functions (IMFs).

Purpose of the Study:

• To introduce a novel, automated procedure for separating noisy IMFs from sEMG signals.
• To enhance the denoising performance of sEMG signals using a new decomposition technique.

Main Methods:

• A novel denoising procedure, Partly EEMD (PEEMD), was developed by integrating Permutation Entropy (PE) into the EEMD sifting process.
• PEEMD automatically separates noisy IMFs based on a predefined PE threshold, reconstructing the denoised signal from selected IMFs.
• The method was validated on experimental sEMG data from eight subjects across six upper limb movement classes, with Sample Entropy (SE) used as a comparative measure.

Main Results:

• The PEEMD denoising procedure demonstrated superior performance compared to traditional EMD, EEMD, and CEEMD methods.
• Quantitative evaluation using Signal to Noise Ratio (SNR), Root Mean Square Error (RMSE), and Reconstruction Error (RE) confirmed the effectiveness of PEEMD.
• Results averaged across subjects showed significant improvements in denoising performance on experimental sEMG data.

Conclusions:

• The proposed PEEMD method offers an effective and automated approach for denoising sEMG signals.
• This technique shows significant potential for improving the accuracy and reliability of sEMG-based applications in muscle diagnosis, rehabilitation, and prosthetics.
• PEEMD outperforms existing state-of-the-art EMD-based denoising techniques for sEMG data.